|Home | About | Journals | Submit | Contact Us | Français|
Persistent infection with high-risk types of human papillomavirus (HPV) is a necessary step in the development of cervical cancer. The incorporation of HPV detection into cervical screening programs may improve the ability to identify women at risk of cervical cancer. We recently evaluated the performance characteristics of a newly developed HPV detection assay, the GenoArray (GA) genotyping assay, for the detection of HPV infections by comparing it with the commercial Roche Linear Array (LA) HPV genotyping assay. The GA assay has an analytical sensitivity for the detection of HPV types 16 (HPV-16) and HPV-18 of as few as 10 to 50 copies, and its reproducibility is adequate. The GA and LA assays showed no significant difference in the rates of detection of genotypes detected by both HPV genotyping assays and oncogenic genotypes, and the interassay agreement was excellent. The GA and LA assays revealed either concordant or compatible genotyping results for 97.5% of the samples and discordant results for only eight (2.5%) samples. Compatible results were also observed for the detection of single or multiple HPV infections and the detection of most of the genotypes. The GA assay also demonstrated good clinical performance characteristics when the comparisons were carried out with clinical subgroups of samples from patients with normal cytologies, low-grade or high-grade squamous intraepithelial lesions, and cancers. Therefore, the GA assay appears to be highly sensitive and specific for the genotyping of HPV. It has the advantage that it specifically detects HPV-52, which overcomes a limitation of the LA assay, and hence, it has potential value for use for genotyping, especially in regions where HPV-52 has a high prevalence.
Human papillomavirus (HPV) is the main etiological agent for the development of cervical cancer. Persistent infection with high-risk (HR) HPV types is recognized as a necessary step in the progression to neoplastic disease (26). Tests for HPV DNA have shown encouraging results when they are used in conjunction with the traditional cytology screening method (3, 7, 8). Many molecular methods for testing for HPV are currently available. The Digene/Hybrid Capture II (HC II) commercial HPV detection kit and the Cervista HPV HR and Cervista 16/18 tests are approved by the FDA for use for routine screening for HPV (9, 19). However, both the HC II and the Cervista HPV HR assays are unable to discriminate specific genotypes or to identify infections involving multiple genotypes, and the Cervista assay detects only two HPV types (types 16 and 18). Recent studies have suggested that different high-risk HPV types have different oncogenic potentials (5). Besides the clinical aspect, HPV genotyping is also a key part of epidemiological studies of HPV infections worldwide.
A number of PCR-based HPV genotyping assays have been developed to amplify HPV DNA, followed by reverse hybridization against immobilized genotype-specific probes, allowing the simultaneous identification of a broad range of anogenital HPV genotypes. However, the different assays have potential variations in their abilities to detect different HPV types due to their different analytical sensitivities and specificities and their failure to detect specific variants (2, 17, 20).
The HPV GenoArray (GA) test (Hybribio Ltd., Hong Kong) is a newly developed PCR-based HPV genotyping assay. It utilizes L1 consensus primers to simultaneously amplify 21 HPV genotypes, followed by flowthrough hybridization with immobilized genotype-specific probes. It is marked as Conformité Européenne (CE) for use in Europe and is currently being used in some hospitals in China.
The present study was designed to evaluate the analytical and clinical performance characteristics of the Hybribio GA test, and the level of concordance of the genotypes detected by the GA test was compared to that of the Roche Linear Array (LA) HPV test. We focused on interassay agreement for overall HPV DNA detection and for the detection of HPV oncogenic risk types. We also assessed the interassay agreement for type-specific as well as single and multiple HPV infections.
One hundred two cervical cancer tissue specimens were retrospectively collected at the Department of Obstetrics and Gynaecology, Queen Mary Hospital, The University of Hong Kong. The ages of the patients ranged from 23 to 89 years, and the median age was 58 years. The cancer tissue specimens were frozen and stored in liquid nitrogen. Histological confirmation and evaluation of the cancer tissue samples were carried out by pathologists to ensure that the samples contained over 70% cancer cells.
Two hundred sixteen Thin-Prep-processed cervical cytology samples, including 24 samples with normal cytologies, 93 samples with low-grade squamous intraepithelial lesions (LSILs), and 99 samples with high-grade squamous intraepithelial lesions (HSILs), were retrieved from the Cervical Cytology Laboratory, Department of Pathology, The University of Hong Kong. The samples had been examined by pathologists for diagnosis, and the reports were reviewed (by A. N. Y. Cheung). The median ages of the subjects with samples with normal cytologies, LSILs, and HSILs were 37 years (range, 22 to 55 years), 37 years (range, 18 to 58 years), and 37 years (range, 16 to 54 years), respectively.
The use of human tissues and cytology specimens was approved by the local institutional ethics committee (Institutional Review Board approval nos. 05-143 T/806, UW 06-362 T/1387, 01-19 RC/B/220, and UW 04-267 T/589).
Cells were collected from the remnants of the ThinPrep cytology samples by centrifugation and were washed with phosphate-buffered saline. Genomic DNA was extracted by proteinase K digestion and the conventional phenol-chloroform-ethanol extraction method (21). A similar protocol was also applied to the extraction of DNA from frozen cervical cancer tissue samples. The quantity of nucleic acid extracted was measured with a spectrophotometer, and the quality was examined by amplification of the human beta-globin gene.
Plasmids containing the full-length genomes of HPV type 16 (HPV-16) and HPV-18 were used for determination of the analytical sensitivity of the GenoArray assay. They were diluted to a series of concentrations of 5 × 103, 1 × 103, 5 × 102, 1 × 102, 5 × 101, 1 × 101, and 5 × 10−1 copies and mixed with 30 ng of C33-A cell genomic DNA (HPV negative). Two different lots of the GenoArray reagent and two PCR machines were used. Each plasmid concentration was tested in triplicate by two operators using two different lots of reagents in two PCR machines.
A panel of 22 cervical swab samples was selected to test the reproducibility of the GenoArray assay. In all, 47 comparisons were made by using 2 samples that were positive for a single HPV type, 18 samples that were positive for various combinations of multiple HPV types (two to three types), and 2 samples that were HPV negative. The 22 members of the test panel were tested in duplicate by using two different lots of reagents.
The GA test is an L1 consensus primer-based PCR assay and is capable of amplifying 21 HPV genotypes, including 13 HR types (types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68), 2 probable HR (PHR) types (types 53 and 66), and 6 low-risk (LR) and unknown-risk (UR) types (types 6, 11, 42, 43, 44, and CP8304 [HPV-81]) (Table (Table1).1). The assay was performed according to the manufacturer's protocol. Briefly, PCR was performed with a reaction volume of 25 μl containing 5 μl of DNA template, 19.25 μl of the master mixture provided, and 0.75 μl of DNA Taq polymerase in a Perkin-Elmer GeneAmp PCR system 9700 apparatus (Applied Biosystems). The amplification protocol was as follows: 9 min of denaturation at 95°C and 40 cycles of 20 s of denaturation at 95°C, 30 s of annealing at 55°C, and 30 s of elongation at 72°C, followed by a final extension for 5 min at 72°C. The amplicon was subsequently denatured and subjected to hybridization. The assay utilized a flowthrough hybridization technique by actively directing the targeting molecules toward the immobilized probes within the membrane fibers, with the complementary molecules being retained by the formation of duplexes. After a stringent wash, the hybrids were detected by the addition of the streptavidin-horseradish peroxidase conjugate (provided with the kit), which binds to the biotinylated PCR products, and a substrate (nitroblue tetrazolium-5-bromo-4-chloro-3-indolylphosphate) to generate a purple precipitate at the probe dot. The results were interpreted by direct visualization. The positive control provided with the kit and the two negative controls (HPV-negative C33-A cells and PCR water) were included in each set of PCRs to assess the performance of the test. After hybridization, the presence of a positive result for both the “internal control” and the “biotin” dots within the membrane indicated that the isolated DNA was of good quality, the enzyme conjugate was valid, and the hybridization process was proper.
The LA test from Roche Diagnostics uses biotinylated PGMY09/11 consensus primers to amplify a 450-bp region of the L1 gene and is capable of detecting 37 HPV genotypes, including 15 HR types (Table (Table1).1). The test was performed according to the manufacturer's instructions. Briefly, DNA was amplified by PCR in a Perkin-Elmer GeneAmp PCR system 9700 apparatus (Applied Biosystems). The denatured PCR product was then hybridized to an array strip containing immobilized oligonucleotide probes. The results were interpreted by using the reference guide and reading the matching individual types down the length of the strip.
HPV type-specific PCR was performed with L1-type-specific primers to clarify the types discrepant between the two genotyping assays. The PCR amplicon was further confirmed by DNA sequencing.
The results obtained by the GA assay were compared with those obtained by the LA assay. Since the GA and LA assays use probes with an overlapping range of HPV genotypes, in order to have an accurate comparison between the two tests, only the HPV genotypes identified by both assays (i.e., HR HPV types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68; PHR types 53 and 66; LR HPV types 6, 11, and 42) were considered in the direct comparison of the individual HPV genotypes detected. These genotypes are referred to as genotypes detected by both assays, or assay-common genotypes. When the results of the GA and LA genotyping assays were compared, the results were classified as concordant, compatible, or discordant on the basis of the following definitions. The results were classified as concordant if the analyses yielded identical assay-common genotypes (single or multiple) or were negative by both assays. If one or more additional assay-common genotypes were detected by either of the assays, the results were classified as compatible. Discordant results indicated no similarity in the detection of the assay-common genotypes by the two assays.
All data were analyzed by using SPSS (version 13.0) statistical software (SPSS Inc., Chicago, IL). The agreement between the results of the two detection methods was assessed by the use of absolute agreement and Cohen's kappa statistics. The crude percent (absolute) agreement between the GA and the LA assays was the percentage of samples with identical results by both methods (16). The unweighted kappa statistics were calculated to determine the level of chance-adjusted agreement between the pair of assay methods (25). A kappa value of 0 indicates no agreement better than chance, and a kappa value of 1 indicates perfect agreement. Kappa values from 0 to 0.20, 0.21 to 0.40, 0.41 to 0.60, 0.61 to 0.80, and above 0.81 indicate poor, fair, moderate, good, and excellent agreement, respectively. The nonparametric McNemar test was used to analyze the complementarities of the detection methods and to determine if the results obtained by the two methods were significantly different. P values of <0.05 were considered statistically significant.
All confirmed sequences of the HPV genotypes (see Table Table5)5) were deposited in the GenBank database under accession numbers GU296023 to GU296094 (see Table S1 in supplemental material).
The analytical sensitivity of the GA assay was determined with plasmids carrying the genome of HPV-16 or HPV-18, as these two HPV types are high-risk types and are the most important oncogenic agents. Two different lots of the GA assay reagent were able to detect as few as 10 to 50 copies of each HPV genome-containing plasmid mixed with genomic DNA. There was a slight difference in the hybridization signals obtained with the two lots of the reagent. One lot of the reagent showed a signal strength less intense than that of the other lot, but the analytical sensitivities were similar by this particular test with the two lots of reagents.
The reproducibility of the GA assay was evaluated by testing a panel of 22 cervical swap samples, which included 188 comparisons (Table (Table2).2). The overall agreement between the replicate results for each sample was 100%. Three single types were shown to be weakly positive with one lot of the reagent but were not detected with the other lot of the reagent. The absolute agreement for the 188 comparisons between the two different lots of the reagent was 93.6%. The discrepant results were mainly due to variations in the hybridization signal strengths between the two lots of reagent used, as the overall signal intensity was slightly stronger with one lot of the test reagent than with the other one.
We first compared the results for the overall detection of HPV DNA by the two methods. The overall detection rate of HPV DNA by the GA assay was 92.5% (294 of 318 samples), whereas it was 95.9% (305 of 318) by the LA assay. The absolute agreement between the GA and the LA assays was 95.9% and the kappa value of 0.63 indicated good agreement (Table (Table3).3). Although the overall rates of HPV positivity by the two methods were significantly different (P = 0.003), this was mainly due to the different numbers of genotypes detected by the two assays (the LA assay detects 37 genotypes, and the GA assay detects 21 genotypes).
For the assay-common HPV genotypes, the overall rates of detection by the two assays were the same (91.8%; 292 of 318). No significant difference was observed between these two genotyping assays (P = 1.0), the absolute agreement was 97.5%, and the kappa value was 0.83, indicating an excellent agreement between the two methods.
For the detection of oncogenic HPV genotypes (13 HR types and 2 PHR types), the GA assay showed a rate of detection similar to that of the LA assay: 90.9% (289 of 318) and 90.6% (288 of 318), respectively. Similar to the results of assay-common HPV genotype detection, no significant difference was observed between the two genotyping assays (P = 1.0) and the interassay agreement was excellent (kappa value = 0.87).
The levels of concordance of genotype-specific detection by the GA and the LA assays were compared: 227 samples (71.4%) showed concordant results and 83 (26.1%) showed compatible results by the two assays; only 8 samples (2.5%) showed discordant genotyping results. Thus, the two genotyping assays revealed either concordant or compatible genotyping results in 97.5% of the samples. Eighty-three samples with compatible genotyping results were further analyzed. For 36 of 83 samples with compatible results, the GA assay identified additional types compared to the types identified by the LA assay. Conversely, the LA assay detected additional types in 39 samples. For the remaining 8 samples, the GA and LA assays identified other HPV genotypes.
Subsequently, a comparison of the specific identification of individual HPV genotypes by the GA and LA genotyping assays was performed, and the results are summarized in Table Table4.4. Neither assay detected HPV-45, probably because of its low prevalence, and the two assays had perfect agreement for the detection of HPV-35 and HPV-6. For the majority of the assay-common HPV genotypes, the results obtained by the two genotyping methods were not significantly different. The strength of the interassay agreement was considered good (kappa values = 0.73 to 0.77) to excellent (kappa values = 0.81 to 0.96), except for HPV types 51, 56, and 42, for which the two assays showed moderate agreement for detection (kappa values = 0.45 to 0.54). The LA assay detected HPV types 39, 51, and 56 in significantly more samples than the GA assay. Since the LA assay is unable to distinguish HPV-52 from other HR types (HPV types 33, 35, and 58), samples that tested positive for HPV-52 but that were negative for HPV types 33, 35, and 58 individually were considered HPV-52 positive. If the samples tested positive for HPV type 33, 35, or 58, they were considered to have an uncertain HPV-52 status and in the present analysis were considered to be HPV-52 negative. Therefore, the GA assay was more sensitive for the detection of HPV-52 than the LA assay, as it was able to identify HPV-52 in the presence of infections with multiple HPV types.
A discrepancy analysis by type-specific PCR followed by sequencing was performed to clarify those discrepant types in the compatible and discordant cases. As we focused on the oncogenic HPV types, 86 cases with 106 discordant types were reassessed. As shown in Table Table5,5, 28 of the 53 (52.8%) discordant types detected by the GA assay were proven to have oncogenic types, while the majority of the discrepant types (44 of 53, 83%) detected by the LA assay were confirmed to have oncogenic types. The LA assay seems to be more sensitive and accurate for the detection of HPV types 16, 31, 39, 51, 56, and 58 and less sensitive for the detection of HPV-33 and HPV-52 than the GA assay.
In the case of single and multiple HPV infections, no significant difference was observed between the GA and the LA assays (P = 1.0) (Table (Table6).6). Both assays detected multiple infections in about 46% of the samples, demonstrating good interassay agreement (kappa value = 0.73). The majority of the infections with multiple HPV types found were double infections: 27.7% (88 of 318) and 28.3% (90 of 318) by the GA and the LA assays, respectively (Table (Table7).7). The highest numbers of HPV types (five types) found in a single sample were identified by GA assay.
The relationship between the concordance of HPV detection by the two detection methods and the cytological and histological diagnoses was analyzed, and the results are shown in Table Table8.8. The comparison was further performed with subgroups of samples with normal or abnormal cytological and histological diagnoses. There was no significant difference between the GA and the LA assays in the detection of assay-common and oncogenic HPV genotypes in subgroups of samples with normal cytologies, LSILs, HSILs, and cancer (P = 1.0). The interassay agreements were good to perfect (kappa values = 0.63 to 1.0). All samples with cervical cancer (n = 102) were found to be positive for the oncogenic HPV types by the two assays, and the results of the two assays were in either complete or partial agreement (i.e., they had concordant and compatible results). Concordant and compatible results were also found for 83.3%, 97.8%, and 98% of the samples in the normal cytology, LSIL, and HSIL groups, respectively (Table (Table99).
The differences in the performance of HPV genotyping tests can strongly affect the results of epidemiological studies and the clinical treatment strategy selected. In the present study, we evaluated a newly developed HPV genotyping test, the GenoArray assay, which has been used clinically in some hospitals in mainland China (14) and other countries, such as Turkey and Israel (10, 13). However, this test has been compared only to the Amplicor HPV test (Roche Diagnostics) in a study with a limited number of samples (13). Although that study demonstrated an excellent agreement between the two tests (kappa value = 0.98) for the overall detection of HPV, other comparisons were not performed, which might be due to the limitation of the Amplicor test (which is not a genotyping test) and the small number of samples tested. In the present study, the analytic and clinical performance characteristics of the GA assay were evaluated, and the level of concordance between the results of the GA assay and the LA HPV assay was compared by testing a large number of samples. The clinical samples tested were selected so that the sample collection contained a range of pathologies, from samples that were cytologically normal to samples that had cervical cancer, in order to test for a broad spectrum of HPV genotypes. Therefore, the results for HPV detection demonstrated in the present study cannot be considered representative of the prevalence of HPV infection in a random population.
The analytical performance of the GA assay was assessed. It had an analytical sensitivity for the detection of HPV types 16 and 18 of as few as 10 to 50 copies. The highly reproducible results were also demonstrated with clinical samples. However, various signal strengths were observed when different lots of the reagent provided with the test kit were used, which accounted for the discrepancy in the genotyping results of up to 6.4% in this study.
The LA assay utilizes the PGMY09/11 PCR primer set, which has commonly been used for the genotyping of HPV in studies of the natural history of HPV (4, 12, 23), and therefore, it served as a reasonable assay for use for comparison with the GA assay in this analysis. The GA assay uses a PCR primer set which is different from PGMY09/11, but it also targets the L1 region of the HPV genome to amplify a broad range of HPV genotypes (21 types). The strength of the interassay agreement increased when the assay-common and oncogenic HPV genotypes detected were compared, and no significant difference in the detection of these HPV genotypes was obtained between the two assays. In general, the GA assay is highly compatible with the commercial LA assay for the detection of assay-common and oncogenic HPV genotypes and single and multiple infections. This was further confirmed when the concordance of the results of the tests was assessed. The GA and the LA assays revealed either concordant or compatible genotyping results for 97.5% of the samples and discordant genotypes for only eight (2.5%) samples. However, the discrepancy analysis revealed differential sensitivities in genotyping between the two assays. The GA assay was less sensitive than the LA assay for the detection of some genotypes. The type-specific effects might partly be attributed to the different primer target sites of some HPV types and an insufficient match between any of the primers in the mixture. The LA assay grossly underestimated the presence of HPV-52, because the assay cannot distinguish it from HPV type 33, 35, or 58 in samples coinfected with those types, which is a known limitation of the assay. This dilemma may limit the application of the LA assay in future epidemiological studies, since HPV-52 is prevalent in approximately 5% of HPV-positive women with normal cervical cytologies (6) and in 8.9 to 15.2% of HPV-positive LSILs and HSILs and 2.2 to 5.2% of cervical cancer cases (15, 18). The GA assay overcomes this limitation and showed a better analytical specificity for the detection of HPV-52. In addition, HPV-52 infection is relatively more common in Asia-Pacific populations than in populations in Western countries (11). It is also one of the five most common high-risk HPV types found in women with normal cytology findings in most of the countries in South Asia (11, 22, 24). Therefore, the GA assay has an advantage for use in the implementation of an HPV screening program and in epidemiological studies in these regions.
To evaluate the clinical performance of the GA assay, the HPV genotyping results of two assays were stratified and compared by the use of samples with normal cytology findings or abnormal cytology and histology diagnoses. Interestingly, there was no significant difference in the detection of assay-common or oncogenic genotypes between the two genotyping assays for the subgroups of samples from women with normal cytology findings, LSILs, HSILs, or cancer. The interassay agreements ranged from good to perfect, indicating the congruent clinical performance of both the GA assay and the LA assay. However, with samples with normal cytology findings, the levels of concordance of the assays was lower (83.3%) than the levels of concordance with samples from the LSIL, HSIL, and cancer groups (97.8 to 100%). This might be due to the small sample size (24 samples) for the group with normal cytology findings. It has also been suggested that women with normal cytologies harbor a large range of HPV genotypes with particularly low viral loads. Therefore, the HPV detection rate in these women may simply reflect the analytical sensitivity of the assay used (1).
In conclusion, the present study demonstrated that the HPV genotyping results obtained by the GA assay are highly compatible with those obtained by the LA assay, with a very small number of samples with discordant results being detected. In general, the two assays are equally easy to perform. The PCR preparation and amplification times are similar. Fifteen and 24 samples can be processed in one hybridization procedure by the GA assay and the LA assay, respectively, but the hybridization time of the GA assay is about one-half that of the LA assay. Therefore, up to 30 or 24 samples can be processed by one operator in 1 working day by the GA assay and the LA assay, respectively. On the other hand, each genotyping assay has its own advantages. The cost of the GA assay is about one-fourth the cost of the LA assay; the GA assay thus has an advantage as an HPV genotyping test for implementation as part of cervical screening programs in developing countries. The LA assay detects more HPV genotypes, while the GA assay is able to detect HPV-52 specifically, which is important for genotyping, especially in regions where HPV-52 has a high prevalence. Therefore, the GA genotyping assay appears to be an accurate and sensitive method for the detection and genotyping of HPV infections, and its potential value needs to be further determined in prospective clinical and epidemiological studies.
This study was jointly funded by the Wong Check She Charitable Foundation and the Government Matching Grant Scheme Funding from the Department of Obstetrics and Gynaecology, The University of Hong Kong.
Published ahead of print on 30 December 2009.
†Supplemental material for this article may be found at http://jcm.asm.org/.